JOURNAL OF GEOPHYSICAL RESEARCH, VOL. 113, D16307, doi:10.

1029/2007JD009362, 2008

Measurement of low-altitude CO over the Indian subcontinent

by MOPITT
J. Kar,1 D. B. A. Jones,1 James R. Drummond,1 J. L. Attié,2 J. Liu,1 J. Zou,1 F. Nichitiu,1
M. D. Seymour,1 D. P. Edwards,3 M. N. Deeter,3 J. C. Gille,3 and A. Richter4
Received 7 September 2007; revised 15 April 2008; accepted 28 April 2008; published 27 August 2008.

[1] We show that the dayside MOPITT retrievals in the lower troposphere can provide
useful information on surface sources of atmospheric CO over the Indian subcontinent. We
find that MOPITT retrievals at 850 hPa show localized enhancements over the Indian
subcontinent, which correlate with similar enhancements seen in the tropospheric NO2
columns from the SCIAMACHY instrument. In particular, high concentrations of CO over
the Indo-Gangetic basin and some prominent cities are captured in the lower-tropospheric
retrievals in spring. MOPITT averaging kernels (normalized to take into account the
absorber amounts in the layers) indicate that the retrievals are sensitive to CO in the lower
troposphere. In winter, MOPITT retrievals at 850 hPa can detect the strongest source areas
over the eastern states of Bihar and West Bengal, thus confirming the so-called ‘‘Bihar
pollution pool,’’ which was detected earlier in the aerosol measurements by the multiangle
imaging spectroradiometer (MISR) aboard Terra. The pollution features are consistent
with the spatial distribution of CO emissions in India, as reflected in the GEOS-Chem
simulation of CO. Furthermore, these lower-tropospheric features in the simulation are
still present after smoothing the modeled fields using the MOPITT averaging kernels and a
priori profile, demonstrating that the retrievals do have sensitivity in the lower
troposphere. This work indicates that although MOPITT retrievals are often most sensitive
to CO in the middle and upper troposphere, they do provide information on lower-
tropospheric CO in selected continental regions with strong thermal contrast and could be
useful for pollution studies.
Citation: Kar, J., et al. (2008), Measurement of low-altitude CO over the Indian subcontinent by MOPITT, J. Geophys. Res., 113,
D16307, doi:10.1029/2007JD009362.

1. Introduction can greatly enhance our knowledge of the targeted species,

long-term changes and trends over a large area can only be
[2] Carbon monoxide (CO) is an important tropospheric
obtained through space-based remote sensing observations.
species inasmuch as it influences the atmospheric chemistry
Such measurements of CO have been made by the Measure-
through its reaction with the OH radical. It is a by-product
ments of Pollution in the Troposphere (MOPITT) since
of combustion and is a useful tracer of atmospheric trans-
2000 [Drummond, 1992]. MOPITT measurements have
port, because of its long lifetime (few weeks). In particular,
been used effectively to study CO plumes from forest fires
emission and subsequent long-range transport of CO from
[Liu et al., 2005], intercontinental transport of pollution
Asia has been the focus of much attention in recent years,
[Edwards et al., 2003, 2006; Heald et al., 2003], frontal
because of strong industrial activities and biofuel burning in
lifting and convective transport of pollution to the upper
this area [Bey et al., 2001; Liu et al., 2003; Heald et al.,
troposphere [Liu et al., 2006; Kar et al., 2004, 2006; Li
2003]. These efforts have relied on a combination of
et al., 2005], and for inverse modeling to quantify surface
ground-based, aircraft, and space-based measurements.
sources of CO [Arellano et al., 2004, 2006; Heald et al.,
While intensive field campaigns using several platforms
2004; Kopacz et al., 2007].
[3] While MOPITT measurements have been useful in
1
Department of Physics, University of Toronto, Toronto, Ontario,
providing a global picture of the distribution of CO, their
Canada. application for near-surface studies has been limited. Since
2
Laboratoire d’Aérologie, Université de Toulouse, CNRS/INSU, the MOPITT operational retrievals employ radiances mea-
Toulouse, France. sured near 4.7 mm, they are sensitive to CO mostly in the
3
Atmospheric Chemistry Division, National Center for Atmospheric
Research, Boulder, Colorado, USA.
middle and upper troposphere with generally less sensitivity
4
Institute of Environmental Physics, University of Bremen, Bremen, in the lower troposphere, as compared to the CO measure-
Germany. ments in the solar channel at 2.3 mm by SCIAMACHY
[Buchwitz et al., 2006, 2007]. It is generally assumed that
Copyright 2008 by the American Geophysical Union. all lower-tropospheric MOPITT retrievals have little sensi-
0148-0227/08/2007JD009362

D16307 1 of 13
D16307 KAR ET AL.: MOPITT CO OVER INDIA
tivity. This perception is due, in part, to the fact that the
primary application of MOPITT has been in the context of
intercontinental transport of pollution [e.g., Edwards et al.,
2003, 2006; Heald et al., 2003; Li et al., 2005], focusing,
for example, on transpacific or transatlantic transport, where
the lower-tropospheric retrievals do indeed have little sen-
sitivity because of the cold ocean. However, Deeter et al.
[2007] recently pointed out that interpretation of the vertical
sensitivity of MOPITT retrievals, as reflected in the aver-
aging kernels, has been influenced by the nonuniform
thickness of the layers of the retrieval grid. They found
that after normalizing for these grid effects, the averaging
kernels indicate useful sensitivity to CO in the lower
troposphere, particularly over dayside tropical regions with
sparse vegetation [Deeter et al., 2007]. In this paper we
explore this issue of the sensitivity of the lower-tropospheric
MOPITT retrievals with emphasis over the Indian subcon-
tinent, where the thermal contrasts are expected to be high.
[4 ] We use lower-tropospheric MOPITT retrievals
along with measurements of tropospheric NO2 from the
SCIAMACHY instrument to assess the sensitivity of the
MOPITT retrievals to the surface sources of CO. We focus
on India because the high population density and recent
accelerated industrial activity imply strong sources of CO
in the region. In addition, while a number of aerosol
studies over India have been carried out in recent years
with emphasis on the Indo Gangetic (IG) basin [e.g., Singh
et al., 2004; Jethva et al., 2005; Tripathi et al., 2006; Nair
et al., 2007], few studies have been published on the
distribution and evolution of the gas species, particularly
using the remotely sensed measurements of the various
tropospheric species that are now available. We examine
here the spatial structures in the distribution of CO in lower-
tropospheric MOPITT retrievals and assess their consistency
with our a priori knowledge of the CO sources in India, as
reflected in the emission inventory in the GEOS-Chem
model and in the model simulation of the horizontal and
vertical structure of CO. We show that the low-altitude
MOPITT retrievals do indeed capture some of the source
information over this region.
2. Data
2.1. MOPITT
[5] The MOPITT instrument is onboard the Terra space-
craft, which is flying in a polar sun synchronous orbit at
an altitude of 705 km. Upwelling infra red radiation near
4.7 mm is measured using gas correlation radiometry which
is then used to retrieve the CO mixing ratios. The CO mixing
ratio profile is reported at 7 levels (surface, 850, 700, 500,
350, 250 and 150 hPa levels) although these measurements
are not independent of one another [Deeter et al., 2004]. The
MOPITT retrieval (^x) can be described as a linear estimate of
the true atmospheric state (xtrue)
^x ¼ xa þ Aðxtrue  xa Þ þ e ð1Þ
where xa is the MOPITT a priori CO profile, A is the
MOPITT averaging kernel matrix, and e is the retrieval
error. The averaging kernel represents the sensitivity of the
MOPITT retrieval to the true state of the atmosphere and
provides a measure of the vertical resolution of the retrieval.
2 of 13
D16307
The trace of the averaging kernel matrix indicates the
number of independent pieces of information (degrees of
freedom for signal) retrieved in the profile. MOPITT profiles
of CO mixing ratios typically have between 1.4– 1.8 degrees
of freedom mostly in the tropics and fall off sharply at higher
latitudes [Deeter et al., 2004].
[6] The MOPITT version 3 retrievals that are used in this
analysis have been validated using in situ aircraft measure-
ments and the results show good agreements with the
average bias being less than 20 ppbv at all levels [Emmons
et al., 2004]. We have used only the dayside retrievals from
MOPITT in this work as these retrievals have the maximum
information content [Deeter et al., 2004].
2.2. SCIAMACHY
[7] The SCIAMACHY instrument was launched on the
European Envisat spacecraft on 1st March 2002 in a sun-
synchronous orbit with an inclination of 98.55° at an
altitude of 800 km [Bovensmann et al., 1999]. The instru-
ment measures backscattered solar radiation in the ultravi-
olet, visible, and near-infrared regions of the spectrum, from
0.24– 2.38 mm, in the nadir and limb modes. The instrument
also makes limb observations using solar and lunar occul-
tation as well as scattered light. In the nadir mode the
horizontal resolution is 30 km  60 km, and the instrument
achieves global observational coverage within 6 days. The
column abundances of a broad range of trace gases have
been retrieved from SCIAMACHY radiances. We use
retrievals of the tropospheric column abundance of NO2
using DOAS algorithm [Richter et al., 2005].
3. Model Description
[8] The GEOS-Chem model is a global three-dimensional
chemical transport model, driven by assimilated meteorol-
ogy from the Goddard Earth Observing System (GEOS-4).
This model has been widely used for various studies of
tropospheric chemistry and transport [e.g., Bey et al., 2001;
Fiore et al., 2003; Heald et al., 2004; Liu et al., 2006;
Zhang et al., 2006]. It has been used previously with
MOPITT data for the study of long-range transport of
pollution [e.g., Heald et al., 2003; Li et al., 2005] and for
inverse modeling of atmospheric CO [e.g., Heald et al.,
2004; Arellano et al., 2004, 2006; Kopacz et al., 2007]. The
version of GEOS-Chem used here is v7-02-04 (http://www-
as.harvard.edu/chemistry/trop/geos/index.html) with a hori-
zontal resolution of 2°  2.5° and with 30 levels in the
vertical from the surface to 0.01 hPa. We simulate the
tropospheric CO distribution for the years 2000– 2004 for
comparison with MOPITT using a linearized version of the
CO chemistry in which monthly mean abundances of OH,
the main sink for CO, are specified from a model simulation
with full nonlinear tropospheric chemistry [Evans and
Jacob, 2005].
[9] The inventory of CO emissions employed in the
model is described in Duncan et al. [2007] with emissions
for the base year of 1985 in the inventory scaled to 1998.
For the period of interest here, May through December, the
total CO emissions in India are 27 Tg CO, with fossil fuel
combustion, biomass burning, and biofuel combustion ac-
counting for 15%, 22%, and 63% of the total, respectively.
The spatial distribution of these emissions is plotted in
D16307 KAR ET AL.: MOPITT CO OVER INDIA D16307

weaker cloud screening employed in the SCIAMACHY

retrievals. Despite these data drop outs, strong spatial
gradients can be seen in the CO 850 hPa distribution. In
particular, there are several areas of localized enhancements
in the CO 850 hPa map, which are also seen in the
corresponding NO2 distribution. Note the enhanced CO
and NO2 values over the IG basin (24– 28°N, 75– 85°E)
as well as over West Bengal and Bangladesh (22– 25°N,
85– 90°E), essentially following the very high population
density in this region. This is quite similar to the recent
GIS based CO Indian national inventory (1°  1°) as well
as higher resolution (0.5°  0.5°) regional emission inven-
tory in Asia (REAS) version 1.1 [Dalvi et al., 2006, their
Figure 6; Ohara et al., 2007]. The MOPITT CO distribution
is also consistent with high anthropogenic pollution levels
along the IG basin observed in the satellite measurements of
aerosol and tropospheric column ozone distributions [Jethva
et al., 2005; Fishman et al., 2003]. Note that there are some
differences in detail between the CO and NO2 distributions,
along the south eastern coast and over the north-eastern
states (20 – 28°N, 92– 96°E). A direct one-to-one correlation

Figure 1. Averaged May– December emissions of CO

used in the GEOS-Chem simulation. Values shown include
the direct emissions of CO from fossil fuel combustion,
biomass burning, biofuel combustion, and the secondary
source of CO from the oxidation of hydrocarbons emitted
by these processes. The units are 1018 molecules cm2
month1.

Figure 1 and shows that the emissions are highest in

northern India, along the IG basin and over the Bihar and
West Bengal region. The studies of Heald et al. [2004] and
Kopacz et al. [2007] conducted inversion analyses for Asian
emissions during February through April 2001, during the
biomass burning season, and found that the model signifi-
cantly overestimated Indian emissions; they found that
biomass burning accounts for 79% of emissions in February
through April. As our focus is on MOPITT data for May –
December, when biomass burning provides a significantly
smaller contribution to the total Indian emissions, the
analysis presented here will not be influenced by this
potential overestimate of biomass burning in the model.

4. Results
4.1. CO Over Indo Gangetic Basin in Spring
[10] Figure 2 shows the distribution of MOPITT CO
mixing ratios at 850 hPa as well as the tropospheric column
NO2 from the SCIAMACHY instrument for May 2004 over
the Indian subcontinent. The data drop outs in the CO
distribution over the southern parts of the country are due to
clouds that can also be seen in the outgoing long-wave
radiation (OLR) maps from NCEP as well as from MODIS
cloud fraction maps during May 2004 (not shown). The
reason why data drop outs are not seen in the corresponding
NO2 distribution could be related to the comparatively
Figure 2. Distribution of MOPITT CO mixing ratios
(ppbv) at 850 hPa (top) and SCIAMACHY tropospheric
column NO2 (1014 cm2; bottom). Data have been binned in
0.5° in latitude and longitude. Both distributions are for
May 2004. Only daytime data from MOPITT measurements
have been used. The white areas indicate data dropouts due
to clouds or topography. Locations of some of the cities
with large population are marked out.

3 of 13
D16307 KAR ET AL.: MOPITT CO OVER INDIA D16307

Figure 3. CO mixing ratio at 850 hPa (ppbv) over the Indian subcontinent in May 2004, as simulated
by GEOS-Chem model after applying the MOPITT averaging kernels (top) and the corresponding
MOPITT CO at 850 hPa (bottom) at the model resolution (2°  2.5°). Note the different scales in the two
cases.

between CO and NO2 may not always be expected because
of differences in the sources and their lifetimes [Buchwitz
et al., 2007]. Over India, domestic biofuel burning domi-
nates the CO emissions [50%, Dalvi et al., 2006], while the
NOx emissions are dominated by fossil fuel burning with
power and transport sectors contributing about 30% each to
the NOx emissions [Garg et al., 2001; Kunhikrishnan et al.,
2006]. Further, nonuniform temporal coverage by the two
instruments and differences in retrieval characteristics asso-
ciated with aerosol interference may also contribute to some
of the differences between the CO and NO2 observations.
[11] In contrast to CO, NO2 is a short-lived species and its
distribution is more indicative of local surface pollution
sources [Richter et al., 2005]. Furthermore, SCIAMACHY
measurements are sensitive to NO2 in the boundary layer
and can clearly localize emissions to individual cities [e.g.,
Buchwitz et al., 2007]. This can also be seen in Figure 2
(bottom) with isolated plumes of NO2 near the locations of
some cities in the subcontinent with large populations and
associated high pollution levels. Interestingly, signatures of
these cities can also be seen in the CO 850 hPa map. Note
the strong isolated plume of CO over Pune, which was
also observed in the CO total column measurements by
SCIAMACHY [Buchwitz et al., 2007]. However, the NO2
plume is located somewhat to the north of the CO plume,
toward Bombay. There are also similar but somewhat
extended plumes of CO around other cities like Delhi,
Lahore and Multan. The CO mixing ratios over Calcutta
and Dhaka are very high and the individual plumes cannot
be distinguished clearly in this color scale. Nonetheless, it
is reasonably clear that the low-altitude MOPITT retrievals
are capturing some of the source information in this case,
which was hitherto thought to be improbable.
[12] The CO distribution at 850 hPa level in May 2004
simulated by the GEOS-Chem model is shown in Figure 3
(top). For comparison, Figure 3 (bottom) shows the

4 of 13
D16307 KAR ET AL.: MOPITT CO OVER INDIA D16307

Figure 4a. MOPITT mean standard (dashed lines) and grid-normalized averaging kernels (solid lines)
for the month of May 2004 over 24– 27°N and 75– 85°E, which includes the Indo Gangetic basin.

MOPITT mixing ratios at 850 hPa gridded to the GEOS- kernels, which are proportionately larger for levels in the
Chem model resolution of 2°  2.5°. MOPITT averaging true profile associated with thicker layers. Grid-normalized
kernels and the a priori have been applied to the model runs averaging kernels take into account the thickness of the
shown in the figure. The two distributions are qualitatively layers (thus the absorber amounts) and these should be used
quite similar, although the model underestimates the CO for a proper interpretation of the averaging kernels [see
mixing ratios. The underestimate is likely due to the CO Deeter et al., 2007 for details]. Figure 4a shows the standard
emission inventory used in the model. In particular, the mean MOPITT averaging kernels for May 2004 over 24–
model bias is largest in northwestern India, where the 27°N and 75– 85°E roughly corresponding to the IG basin
specified emissions in the model are low (Figure 1). Beig area, as well as the grid-normalized kernels for the surface,
and Brasseur [2006], for example, have estimated an the 850 hPa and the 250 hPa levels. Firstly note that the
increase of 10– 20% in CO mixing ratios near the IG plains averaging kernels for both the surface and 850 hPa are
between 1991 and 2001 for the month of July. However, the nonzero at the surface thus indicating sensitivity to the
modeled CO spatial distribution is quite similar to that surface sources of CO. Secondly, the normalized kernel
simulated for May 2001 by Beig and Ali [2006] using the
MOZART-2 chemical transport model with the recent
Indian national inventory for CO emissions [Dalvi et al.,
2006]. As our focus here is primarily to demonstrate the
sensitivity of MOPITT retrievals to CO in the lower
troposphere, we do not attempt to optimize the model
emissions based on the MOPITT observations. However,
the consistency between spatial distributions of CO and
NO2 suggests that a coupled, multispecies inversion analy-
sis of CO and NO2, accounting for the different emission
factors for these species, from biomass burning and fuel
combustion, may provide valuable information to help
independently quantify the CO emissions from these differ-
ent source types.
[13] The MOPITT retrievals are performed with a uni-
form a priori profile, which has a maximum concentration
of CO of 120 ppbv at the surface. The significant departure
of the retrievals shown in Figure 2 from the a priori values
indicate significant sensitivity to lower atmospheric CO in
the retrievals for this region. The sensitivity of the MOPITT
retrievals to CO in the lower troposphere can be assessed by
examination of the MOPITT averaging kernels. However, as Figure 4b. Distribution of the surface value of MOPITT
pointed out by Deeter et al. [2007], the nonuniform spacing grid-normalized averaging kernel for 850 hPa for the month
between the retrieval levels (thus layers with nonuniform of May 2004. Data binned in 0.5° in latitude and longitude
thickness) affects physical interpretation of the averaging and white areas indicate missing data.

5 of 13
D16307 KAR ET AL.: MOPITT CO OVER INDIA D16307

Figure 5. Thermal contrast, defined as the difference between the surface skin temperature (retrieved by
MOPITT) and air temperature at 925 hPa from NCEP reanalyses data for May 2004. Data binned in 0.5°
in latitude and longitude and white areas indicate missing data.

for the surface shows nearly uniform sensitivity to layers
from the surface to 500 hPa, in contrast to the standard
kernel, which shows far higher sensitivity to 700 hPa level
as compared to the surface. Similarly the normalized aver-
aging kernel for 850 hPa also shows enhanced sensitivity to
the surface as compared to the standard kernel.
[14] The distribution of the grid-normalized 850 hPa
averaging kernel evaluated at the surface over the Indian
subcontinent in May 2004 is shown in Figure 4b. This
quantity indicates the sensitivity of the retrieved amount at
the 850 hPa level to perturbations of the absorber amount of
the surface layer in the true profile. Significant sensitivity to
the boundary layer pollution can be seen over most parts of
the country with the highest sensitivity occurring over the
deserts of Rajasthan (24– 30°N, 68– 75°E) in the west.
Interestingly a pronounced minimum sensitivity can be seen
over the north-eastern states between 90– 96°E, which could
be due to the presence of significant forested areas there.
Deeter et al. [2007] found similar low sensitivity over the
Congo and Amazon basins and attributed this to evapora-
tion and evapotranspiration processes.
[15] The sensitivity of the MOPITT retrievals to lower-
tropospheric CO should be related to the thermal contrast
between the surface and the lowest layers of the atmosphere.
Higher thermal contrast yields higher sensitivity to lower-
tropospheric CO. Figure 5 shows the difference between the
surface skin temperature, as retrieved by MOPITT and the
atmospheric temperature at 925 hPa level, as obtained from
the NCEP reanalyses for May 2004. In general, the tem-
perature contrast is quite significant over most of the region
with maximum values of 20 K over the western parts of
the subcontinent. Also, the thermal contrast map is gener-
ally consistent with the grid-normalized 850 hPa averaging
kernel map (Figure 4b), as might be expected. Although the
thermal contrast should be estimated between the surface
and a layer immediately higher up, we have used the NCEP
temperature at 925 hPa to avoid data uncertainties associ-
ated with topographical variations at lower levels. Further,
the retrieval errors and the percent a priori diagnostic at
850 hPa level also show similar distributions over the
country with lower values in the western parts with larger
thermal contrasts (not shown).
[16] The likelihood of directly sensing the surface sources
of CO in May is further increased by the large planetary
boundary layer (PBL) heights over India at this time. The
PBL heights obtained from the European Center for Medium-
range Weather Forecasts (ECMWF) data for 6 UT corre-
sponding to 11:30 am local time which is quite close to the
MOPITT overpass time (10:30 am) indicate PBL heights in
excess of 2 km over much of the central part of the country
and reaching 3 km or more over the IG basin (not shown). It
should be noted that the CO mixing ratio retrieved by
MOPITT would depend upon both the sensitivity as well as
the strength of the source itself. In particular, over the IG
basin, the sensitivity of MOPITT retrievals to emissions in
the surface layer is lower than over the western regions and

6 of 13
D16307 KAR ET AL.: MOPITT CO OVER INDIA D16307

yet the retrievals provide information about the sources

Figure 6. Comparison of MOPITT and GEOS-Chem
profiles of CO on 29 May 2004 at 87°E, 24°N. The
MOPITT profiles have been averaged on the GEOS-Chem
2°  2.5° grid. The retrieved MOPITT profile and the a
priori profile used in the retrieval are shown as the red
dashed and black dotted lines, respectively. The GEOS-
Chem profile of CO is shown as the green solid line,
whereas the modeled profile transformed by the MOPITT
averaging kernels and a priori profile is indicated by the
blue dash-dotted line.
because of the high source strength over the basin.
[17] The vertical structure of the MOPITT and modeled
profiles over eastern India (24°N, 87°E) on 29 May 2004 is
shown in Figure 6. Because of the strong surface source of
CO in the region, the modeled profile decreases from a
maximum CO abundance of about 230 ppbv at the surface
to background values at about 600 hPa. Above 600 hPa, the
modeled CO profile is similar to the MOPITT a priori
profile. Transforming the model profile with the MOPITT
averaging kernels and a priori profile smoothes the modeled
profile. However, the smoothed model profile still shows
significantly higher abundances of CO in the lower tropo-
sphere than in the middle and upper troposphere, due to the
sensitivity to lower troposphere in the retrievals as reflected
in the averaging kernels.
[18] Figure 7a shows the distribution of CO mixing ratios
at 850 hPa averaged over the month of May for the years
2000, 2002, 2003, 2005, 2006, and 2007. Data for May
2001 are not available because of instrument issues
[Emmons et al., 2004]. The strong enhancement over the
IG basin is very clearly observed for each year with
significant interannual variations. This indicates the persis-
tent nature of this pollution feature and shows that MOPITT
is consistently capturing source information. However, the
signature of the various cities in the form of isolated plumes

Figure 7a. CO mixing ratio (ppbv) retrieved at 850 hPa level by MOPITT over the Indian subcontinent
for May 2000, 2002, 2003, 2005, 2006, and 2007. Data binned in 0.5° in latitude and longitude, and
white areas indicate missing data. Only daytime data have been used. Data for May 2001 are not
available.

7 of 13
D16307 KAR ET AL.: MOPITT CO OVER INDIA
Figure 7b. Modeled CO mixing ratios for May 2002
convolved with the corresponding MOPITT averaging
kernels (top), for May 2003 convolved with the correspond-
ing MOPITT kernels (middle), and for May 2003, but
convolved with the MOPITT kernels for May 2002 (bottom).
is not seen uniformly from year to year for the same month,
with the plumes associated with specific cities showing up
strongly in some years and not in others. This might be
related to the local meteorology and data coverage issues
and needs to be further explored.
[19] The inter annual variation is strongest in the early
part of the decade with distinct lower values in 2000 and
2002 as compared to 2003 – 2007, the latter years showing
much less variability. We have investigated the possible
reasons for this variability by focusing on the two months of
May 2002 and May 2003; MOPITT observations in 2003
show significantly more CO than in 2002. We compared the
GEOS-Chem simulation of CO for May 2002 and 2003,
using the same emissions in the model for both years. As
shown in Figure 7b, the modeled CO abundance, after
smoothing with the MOPITT averaging kernels was signif-
icantly greater in May 2003 than in May 2002. To assess the
potential impact of differences in the MOPITT observational
coverage and vertical sensitivity on the modeled CO for
2002 and 2003, we sampled the May 2003 modeled
8 of 13
D16307
distribution based on the observational coverage for May
2002 and applied the 2002 averaging kernels to smooth the
fields. This resulted in lower abundances of CO (Figure 7b,
bottom) than that obtained with the 2003 averaging kernels
(Figure 7b, middle), but more than that produced with the
modeled data and averaging kernel for 2002 (Figure 7b,
top). The results in Figure 7b show that the differences
between May 2002 and 2003 reflect a combination of
differences in the vertical sensitivity and spatiotemporal
sampling of the MOPITT retrievals and in atmospheric
transport of CO due to the different meteorological con-
ditions. Properly quantifying the relative contributions of
these different factors to the year to year variability shown
in Figure 7a will require a more detailed modeling analysis,
which is beyond the scope of this paper.
[20] A noticeable difference between the GEOS-Chem
meteorological fields for May 2002 and 2003 is that the
mean PBL height, averaged over the area defined by 20–
30°N and 68– 88°E, was about 400 m higher in May 2003
(with a height of 2.25 km) as compared to May 2002 (with a
height of 1.84 km). An increased PBL height should mix the
boundary layer pollution higher up where MOPITT has
better sensitivity and may explain the larger variation than
simulated in the model, with constant emission. Further, the
cloud fraction also showed a significant decrease from 0.43
in May 2002 to 0.27 in May 2003. This can also be seen in
the enhanced MOPITT data dropouts in May 2002
(Figure 7a). This non-uniform sampling may also contribute
to the differences.
4.2. ‘‘Bihar Pollution Pool’’ in CO and NO2 in Winter
[21] Di Girolamo et al. [2004] had earlier observed high
aerosol optical depths over the eastern states of Bihar and
West Bengal in winter (the so-called ‘‘Bihar pollution
pool’’) from the data obtained by the MISR instrument,
also on Terra. As they pointed out, strong subsidence and
light winds over this area in winter prevent venting of the
boundary layer pollution to higher altitudes, accumulating
and limiting it close to the source locations. Therefore it
would be interesting to explore the sensitivity of MOPITT
to low-altitude pollution in this area in winter.
[22] Figure 8 (top) shows the MOPITT CO mixing ratios
retrieved at 850 hPa level for the months of October,
November, and December 2004. The corresponding tropo-
spheric columns of NO2 retrieved from SCIAMACHY are
shown (bottom). In October, relatively enhanced CO emis-
sions can still be seen along the IG basin, although at a
much lower level as compared to the month of May. The
high CO mixing ratios seen below 20°N and between 80–
100°E are likely the emissions from Southeast Asia and
China which are transported to the Indian ocean. Note that
there are no MOPITT data over much of the IG basin region
for November and December. This has been observed in the
MOPITT data consistently for all years for the months of
November –January. The reason for this is not clear at this
time but could be related to the intense haze and foggy
conditions that develop each winter over this area [Tripathi
et al., 2006; Gautam et al., 2007] that might be affecting the
CO retrievals, although thick fog layers should also affect
NO2 retrievals. However, the evolution of CO in the lower
troposphere in the rest of the subcontinent can still be
studied. In particular note the large plume of CO that
D16307 KAR ET AL.: MOPITT CO OVER INDIA D16307

Figure 8. MOPITT CO mixing ratios at 850 hPa (ppbv, top) and SCIAMACHY tropospheric NO2
columns (1014 cm2, bottom) for the months of October – December 2004. Data binned in 0.5° in latitude
and longitude and white areas indicate missing data.

develops over the states of Bihar and West Bengal (20–
25°N, 84 – 88°E) in November and further increases in
December. This can be seen in the corresponding NO2
distribution as well. There is no biomass burning in this
region in winter [Di Girolamo et al., 2004]. Therefore this
CO is likely to be the locally generated pollution from
mostly fuel combustion. Thus both the MOPITT CO data at
850 hPa as well as the NO2 tropospheric columns confirm
the Bihar pollution plume during winter over this area in the
gas species, which is remarkably similar to the aerosol
climatology of Di Girolamo et al. [2004] (their Figure 3),
clearly indicating the concentration of pollution which
should have important implications for atmospheric chem-
istry and climate in this region.
[23] Figure 9 shows the corresponding GEOS-Chem
simulations for October – December 2004 at the 850 hPa
level after incorporating the MOPITT averaging kernels
(top). To facilitate comparison between the model and the
data we have replotted the MOPITT 850 hPa mixing ratios
for the 3 months at the GEOS-Chem resolution grid
(bottom). The evolution of the CO distribution in the lower
troposphere as seen in MOPITT retrievals is quite similar to
the model results with the plume over the eastern states of
Bihar and West Bengal reaching a maximum in December.
Analysis of the GEOS-Chem model simulations in which
we tagged the CO emissions from different geographical
regions show that the CO observed over the Bay of Bengal
in the model is from emissions in East Asia and Southeast
Asia (not shown). The modeled CO mixing ratios are lower
than the observed values across India and could be due to an
underestimate of Indian emissions.
[24] The subsidence in the region is reflected in the
vertical profiles of CO simulated by the GEOS-Chem model,
shown in Figure 10. The profiles show high abundances of
CO in the boundary layer, exceeding 250 ppbv, and low
concentrations in the free troposphere, less than about
80 ppbv, which is characteristic of background concentra-
tions of CO in the northern hemisphere. The midtropospheric
concentrations of CO in the model in this region are in fact
less than the MOPITTa priori. For the scene shown in
Figure 10a, the boundary layer is capped at about 800 hPa,
whereas in Figure 10b the boundary layer is capped at about
700 hPa. Transforming the modeled profiles using the
MOPITT averaging kernel and the a priori profile signifi-
cantly smoothes the profiles because of the coarse vertical
resolution of the retrievals. However, the significant depar-
ture of the smoothed model profile from the a priori in the
lower troposphere indicates sensitivity to lower-tropospheric
CO, as reflected in the MOPITT averaging kernels.
[25] The vertical structure of CO in the model depicted in
Figure 10, with low concentrations of CO in the free
troposphere, is indicative of the distribution of CO across
the region. This suggests that the structures in the CO
distribution that are present in the modeled fields at the
850 hPa level after application of the MOPITT averaging
kernel and a priori profile are not due to the influence of CO

9 of 13
D16307 KAR ET AL.: MOPITT CO OVER INDIA D16307

Figure 9. CO mixing ratios at 850 hPa (ppbv) from October to December 2004 as simulated by GEOS-
Chem after application of the MOPITT averaging kernels for each month (top). The corresponding
MOPITT data at the model resolution grid (2°  2.5°) are shown (bottom). Missing data are shown in
white.

in the middle and upper troposphere. Comparison of the
smoothed model and MOPITT profiles in Figure 10 indicate
that the model does provide an underestimate of the middle
and upper tropospheric CO. The bias in the model is larger
in the upper troposphere than in the lower troposphere.
Indeed, for the observations in Figure 10b, the model is
biased slightly high relative to MOPITT in the lower
troposphere, whereas it is biased low in the upper tropo-
sphere. Because the averaging kernels are broad (not
shown), with the greatest sensitivity in the free troposphere,
the presence of a plume of CO in the middle and upper
troposphere will influence the retrieved CO in the lower
troposphere. However, to produce the large departures of
the retrieved profile from the a priori in the lower tropo-
sphere, as shown in Figure 10, requires the presence of a
large source of CO in the lower troposphere. We use the
model here as a proxy for the vertical structure of the true
atmosphere, and although the model simulation does pro-
vide an underestimate of the CO abundance in the middle
and upper troposphere, and therefore an underestimate of
the influence of CO at these levels to the lower-tropospheric
retrievals, the agreement in the vertical structure of the
smoothed model profiles and the retrievals does demon-
strate that the MOPITT retrievals do have sensitivity to
lower-tropospheric CO.
[26] The distribution of the grid-normalized 850 hPa
averaging kernels evaluated at the surface for the month
of December 2004 is shown in Figure 11. While the
sensitivity to the lower-tropospheric pollution has decreased
somewhat in the western region as compared to May, it is
still quite significant over most of the country, particularly
over the Bihar pollution pool region. The latter, coupled
with the strong and trapped emission makes it possible for
MOPITT to detect the Bihar pollution pool in winter. The
Bihar pollution pool was observed in MOPITT data for
years other than 2004 as well, as can be seen in Figure 12,
showing the CO distribution at 850 hPa for 2000, 2001,
2002, 2003, 2005, and 2006, averaged for the month of
December. Once again strong interannual variability can be
seen in the December maps. Tagged CO simulation runs
indicate contributions from the South East Asia and East
Asian source regions during December, variations of which
tend to control the interannual variability of CO seen in the
south-central regions of the country.
[27] Note that in the December CO maps (Figures 8 and
12), apart from the Bihar pollution plume, the spatial
gradients are not as strong as in May. In particular, the
cities are not easily distinguishable in the CO 850 hPa map
as such. This could be due to the increased lifetime of CO in
winter making it more susceptible to transport effects. In
fact, pollutants from the southern areas are transported out
to the Arabian sea by moderately high easterly winds [Di
Girolamo et al., 2004] as indeed can be seen in the
December CO maps for various years. Similar plumes of

10 of 13
D16307 KAR ET AL.: MOPITT CO OVER INDIA D16307

Figure 10. Comparison of MOPITT and GEOS-Chem profiles of CO on 14 December 2004 at (a) 84°E,
20°N and (b) 88°E, 24°N. The MOPITT profiles have been averaged on the GEOS-Chem 2°  2.5° grid.
The retrieved MOPITT profile and the a priori profile used in the retrieval are shown as the red dashed
and black dotted lines, respectively. The GEOS-Chem profile of CO is shown as the green solid line,
whereas the modeled profile transformed by the MOPITT averaging kernels and a priori profile is
indicated by the blue dash-dotted line.

advected CO can also be seen over the Bay of Bengal,
particularly for the years 2003, 2005, and 2006.
[28] Similar strong and localized enhancements in CO
mixing ratios at 850 hPa were seen for other months as well
(not shown), including signature of strong biomass burning
emissions in March as also for all the years with available
data. It is clear that the MOPITT retrievals of low-altitude
CO are able to capture at least part of the lower-tropospheric
pollution particularly where the sources are the strongest.
This is consistent with the work of Deeter et al. [2007],
even though the total CO columns might give quantitatively
lower estimates over these areas as compared to SCIA-
MACHY, the latter having better sensitivity to the boundary
layer [Buchwitz et al., 2006, 2007].
[30] The sensitivity of the MOPITT retrievals in this
region to lower-tropospheric CO was further demonstrated
using vertical profiles of CO from the GEOS-Chem simu-
lation in December when the CO emissions in the model are
trapped in the lower troposphere as a result of atmospheric
subsidence. We showed that the enhanced CO in the lower
troposphere is still present in the modeled fields after
smoothing with the MOPITT averaging kernels. This study
indicates that, MOPITT retrievals may provide useful in-
formation on CO abundances in the lower troposphere, and
together with complementary satellite data, could be used

5. Conclusions
[29] We have presented evidence that MOPITT dayside
retrievals in the lower troposphere over continental areas
with strong thermal contrast are able to provide useful
information on surface emissions of CO. In particular over
the Indian subcontinent, the low-altitude retrievals can
effectively delineate the strong surface source regions over
the IG basin in spring as well as signatures of some large
cities. In winter, the evolution of the so-called Bihar
pollution pool was also clearly visible in the MOPITT
850 hPa retrievals. The large scale pollution features seen
in the CO distribution were also seen in tropospheric
column NO2 data from SCIAMACHY, thus indicating that
MOPITT low-altitude retrievals were indeed capturing the
source regions. These pollution features were shown to be
consistent with a priori knowledge of the spatial distribu-
tion of CO emissions in India as reflected in the GEOS-
Chem simulation of CO over this region, although the
model provided an underestimate of the observed CO
abundance. Figure 11. Surface value of the 850-hPa grid normalized
averaging kernel for December 2004. Data binned in 0.5° in
latitude and longitude and white areas indicate missing data.

11 of 13
D16307 KAR ET AL.: MOPITT CO OVER INDIA D16307

Figure 12. CO mixing ratio (ppbv) retrieved at 850 hPa level by MOPITT over the Indian subcontinent
for December 2000, 2001, 2002, 2003, 2005, and 2006. Only daytime data have been used. Data binned
in 0.5° in latitude and longitude and white areas indicate missing data.

for air quality studies. In particular, the evolution of regional
pollution features like the Bihar pollution pool can be
studied and monitored, since the temporal coverage of
MOPITT now extends over 7 years. Work is needed to
extend this analysis to other high-pollution regions in other
parts of the globe.
[31] Acknowledgments. MOPITT mission and data analysis are
supported by the Canadian Space Agency, the Natural Sciences and
Engineering Research Council, and the National Aeronautics and Space
Administration. D.P.E., M.N.D., and J.C.G. are supported by the NASA
EOS program. J.L.A. is supported by CNES and Région Midi Pyrénées.
The GEOS-Chem model is maintained at Harvard University with support
from the NASA Atmospheric Chemistry Modeling and Analysis Program.
The referees are thanked for the very useful comments which helped
improve the quality of the paper.
References
Arellano, A. F., Jr., et al. (2004), Top-down estimates of global CO sources
using MOPITT measurements, Geophys. Res. Lett., 31, L01104,
doi:10.1029/2003GL018609.
Arellano, A. F., Jr., et al. (2006), Time-dependent inversion estimates of
global biomass-burning CO emissions using Measurement of Pollution in
the Troposphere (MOPITT) measurements, J. Geophys. Res., 111,
D09303, doi:10.1029/2005JD006613.
Beig, G., and K. Ali (2006), Behavior of boundary layer ozone and its
precursors over a great alluvial plain of the world: Indo-Gangetic plains,
Geophys. Res. Lett., 33, L24813, doi:10.1029/2006GL028352.
Beig, G., and G. P. Brasseur (2006), Influence of anthropogenic emissions
on tropospheric ozone and its precursors over the Indian tropical region
during a monsoon, Geophys. Res. Lett., 33, L07808, doi:10.1029/
2005GL024949.
Bey, I., D. J. Jacob, J. A. Logan, and R. M. Yantosca (2001), Asian
chemical outflow to the Pacific in spring: Origins, pathways, and budgets,
J. Geophys. Res., 106, 23,097 – 23,113.
Bovensmann, H., et al. (1999), SCIAMACHY- Mission objectives and
measurement modes, J. Atmos. Sci., 56, 121 – 150.
Buchwitz, M., et al. (2006), Atmospheric carbon gases retrieved from SCIA-
MACHY by WFM-DOAS: version 0.5 CO and CH4 and impact of calibra-
tion improvements on CO2 retrieval, Atmos. Chem. Phys., 6, 2727 – 2751.
Buchwitz, M., et al. (2007), Three years of global carbon monoxide from
SCIAMACHY: comparison with MOPITT and first results related to the
detection of enhanced CO over cities, Atmos. Chem. Phys., 7, 2399 – 2411.
Dalvi, M., G. Beig, U. Patil, A. Kaginalkar, C. Sharma, and A. P. Mitra
(2006), A GIS based methodology for gridding large scale emission
inventory: Application to carbon-monoxide emissions over India region,
Atmos. Environ., 40, 2995 – 3007.
Deeter, M. N., et al. (2004), Vertical resolution and information content of
CO profiles retrieved by MOPITT, Geophys. Res. Lett., 31, L15112,
doi:10.1029/2004GL020235.
Deeter, M. N., D. P. Edwards, J. C. Gille, and J. R. Drummond (2007),
Sensitivity of MOPITT observations to carbon monoxide in the lower
troposphere, J. Geophys. Res., 112, D24306, doi:10.1029/2007JD008929.
Di Girolamo, L., et al. (2004), Analysis of multi-angle imaging spectroradi-
ometer (MISR) aerosol optical depths over greater India during winter
2001 – 2004, Geophys. Res. Lett., 31, L23115, doi:10.1029/2004GL021273.
Drummond, J. R. (1992), Measurements of pollution in the troposphere, in
The Use of EOS for Studies of Atmospheric Physics, edited by J. C. Gille
and G. Visconti, pp. 77 – 101, North-Holland, Amsterdam.
Duncan, B. N., et al. (2007), Global budget of CO, 1988 – 1997: Source
estimates and validation with a global model, J. Geophys. Res., 112,
D22301, doi:10.1029/2007JD008459.
Edwards, D. P., et al. (2003), Tropospheric ozone over the tropical Atlantic:
A satellite perspective, J. Geophys. Res., 108(D8), 4237, doi:10.1029/
2002JD002927.
Edwards, D. P., et al. (2006), Satellite-observed pollution from Southern
Hemisphere biomass burning, J. Geophys. Res., 112, D14312,
doi:10.1029/2005JD006655.

12 of 13
D16307 KAR ET AL.: MOPITT CO OVER INDIA
Emmons, L. K., et al. (2004), Validation of MOPITT CO retrievals with
aircraft in situ profiles, J. Geophys. Res., 109, D03309, doi:10.1029/
2003JD004101.
Evans, M. J., and D. J. Jacob (2005), Impact of new laboratory studies of
N2O5 hydrolysis on global model budgets of tropospheric nitrogen oxi-
des, ozone and OH, Geophys. Res. Lett., 32, L09813, doi:10.1029/
2005GL022469.
Fiore, A. M., D. J. Jacob, R. Mathur, and R. V. Martin (2003), Application
of empirical orthogonal functions to evaluate ozone simulations with
regional and global models, J. Geophys. Res., 108(D14), 4431,
doi:10.1029/2002JD003151.
Fishman, J., A. E. Wozniak, and J. K. Creilson (2003), Global distribution
of tropospheric ozone from satellite measurements using the empirically
corrected tropospehric ozone residual technique: Identification of the
regional aspects of air pollution, Atmos. Chem. Phys., 3, 893 – 907.
Garg, A., P. R. Shukla, S. Bhattacharya, and V. K. Dadhwal (2001), Sub-
region (district) and sector level SO2 and NOx emissions for India: As-
sessment of inventories and mitigation flexibilities, Atmos. Environ., 35,
703 – 713.
Gautam, R., N. C. Hsu, M. Kafatos, and S.-C. Tsay (2007), Influences of
winter haze on fog/low cloud over the Indo-Gangetic plains, J. Geophys.
Res., 112, D05207, doi:10.1029/2005JD007036.
Heald, C. L., et al. (2003), Asian outflow and transpacific transport of carbon
monoxide and ozone pollution: An integrated satellite, aircraft and model
perspective, J. Geophys. Res., 108(D24), 4804, doi:10.1029/2003JD003507.
Heald, C. L., D. J. Jacob, D. B. A. Jones, P. I. Palmer, J. A. Logan, D. G.
Streets, G. W. Sachse, J. C. Gille, R. N. Hoffman, and T. Nehrkorn
(2004), Comparative inverse analysis of satellite (MOPITT) and aircraft
(TRACE-P) observations to estimate Asian sources of carbon monoxide,
J. Geophys. Res., 109, D23306, doi:10.1029/2004JD005185.
Jethva, H., S. K. Satheesh, and J. Srinivasan (2005), Seasonal variability of
aerosols over the Indo-Gangetic plains, J. Geophys. Res., 110, D21204,
doi:10.1029/2005JD005938.
Kar, J., et al. (2004), Evidence of vertical transport of carbon monoxide
from Measurements of Pollution in the Troposphere (MOPITT), Geo-
phys. Res. Lett., 31, L23105, doi:10.1029/2004GL021128.
Kar, J., J. R. Drummond, D. B. A. Jones, J. Liu, F. Nichitiu, J. Zou, J. C. Gille,
D. P. Edwards, and M. N. Deeter (2006), Carbon monoxide (CO) maximum
over the Zagros mountains in the Middle East: Signature of mountain vent-
ing?, Geophys. Res. Lett., 33, L15819, doi:10.1029/2006GL026231.
Kopacz, M., D. J. Jacob, D. K. Henze, C. L. Heald, D. G. Streets, and
Q. Zhang (2007), Comparison of adjoint and analytical Bayesian inversion
methods for constraining Asian sources of carbon monoxide using satellite
(MOPITT) measurements of CO columns, J. Geophys. Res., 113(D7),
D07302, doi:10.1029/2007JD009264.
Kunhikrishnan, T., et al. (2006), Regional NOx emission strength for the
Indian subcontinent and the impact of emissions from India and neigh-
boring countries on regional O 3 chemistry, J. Geophys. Res., 111,
D15301, doi:10.1029/2005JD006036.
13 of 13
D16307
Li, Q., D. J. Jacob, R. Park, Y. Wang, C. L. Heald, R. Hudman, R. M.
Yantosca, R. V. Martin, and M. Evans (2005), North American pollution
outflow and the trapping of convectively lifted pollution by upper-level
anticyclone, J. Geophys. Res., 110, D10301, doi:10.1029/2004JD005039.
Liu, H., D. J. Jacob, I. Bey, R. M. Yantosca, B. N. Duncan, and G. W.
Sachse (2003), Transport pathways for Asian pollution outflow over the
Pacific: Interannual and seasonal variations, J. Geophys. Res., 108(D20),
8786, doi:10.1029/2002JD003102.
Liu, J., J. R. Drummond, Q. Li, J. C. Gille, and D. C. Ziskin (2005),
Satellite mapping of CO emission from forest fires in northwest America
using MOPITT measurements, Remote Sens. Environ., 95, 502 – 516.
Liu, J., et al. (2006), Large horizontal gradients in atmospheric CO at the
synoptic scale as seen by spaceborne Measurements of Pollution in the
Troposphere, J. Geophys. Res., 111, D02306, doi:10.1029/2005JD006076.
Nair, V. S., et al. (2007), Wintertime aerosol characteristics over the IG
Plain (IGP): Impacts of local boundary layer processes and long-range
transport, J. Geophys. Res., 112, D13205, doi:10.1029/2006JD008099.
Ohara, T., H. Akimoto, J. Kurokawa, N. Horii, K. Yamaji, X. Yan, and
T. Hayasaka (2007), An Asian emission inventory of anthropogenic
emission sources for the period 1980 – 2020, Atmos. Chem. Phys., 7,
4419 – 4444.
Richter, A., J. P. Burrows, H. Nüß, C. Granier, and U. Niemeier (2005),
Increase in tropospheric nitrogen dioxide over China observed from
space, Nature, 437, 129 – 132, doi:10.1038/nature04092.
Singh, R. P., S. Dey, S. N. Tripathi, V. Tare, and B. Holben (2004), Varia-
bility of aerosol parameters over Kanpur, northern India, J. Geophys.
Res., 109, D23206, doi:10.1029/2004JD004966.
Tripathi, S. N., et al. (2006), Measurements of atmospheric parameters
during Indian Space Research Organization Geosphere Biosphere Pro-
gramme Land Campaign II at a typical location in the Ganga basin: 1.
Physical and optical properties, J. Geophys. Res., 111, D23209,
doi:10.1029/2006JD007278.
Zhang, L. D., et al. (2006), Ozone-CO correlations determined by the TES
satellite instrument in continental outflow regions, Geophys. Res. Lett.,
33, L18804, doi:10.1029/2006GL026399.

J. L. Attié, Laboratoire d’Aérologie, Université de Toulouse, CNRS/
INSU, 14 Avenue Edouard Belin, 31400 Toulouse, France.
M. N. Deeter, D. P. Edwards, and J. C. Gille, Atmospheric Chemistry
Division, National Center for Atmospheric Research, P.O. Box 3000,
Boulder, CO 80307, USA.
J. R. Drummond, D. B. A. Jones, J. Kar, J. Liu, F. Nichitiu, M. D.
Seymour, and J. Zou, Department of Physics, University of Toronto, 60 St.
George Street, Toronto, ON, Canada M5S 1A7. (jkar@atmosp.physics.
utoronto.ca)
A. Richter, Institute of Environmental Physics, University of Bremen,
Otto-Hann-Allee 1, D-28359 Bremen, Germany.